I think lzc shouldn't be using any caches at all.  Caches are for servers.  
They make no sense for command-line invocation IMO.

On 2011-02-03, at 14:39, Henry Minsky wrote:

> That does look bogus; it's only for the command line invocation of lzc, but
> the code is just appending the java tmpdir with the directory name
> "lzccache", so it clearly will conflict when different versions of lzc from
> different workspaces get called.
> 
> The server uses WEB-INF/lps/work/ for it's caches, I guess lzc should be
> doing that as well.
> 
> I had better check that whatever locking we're using for the server also
> applies when
> called from lzc...
> 
> 
> 
> 
> On Thu, Feb 3, 2011 at 1:48 PM, P T Withington <[email protected]>wrote:
> 
>> Not approved.
>> 
>> There is actually a deeper problem here.
>> 
>> Test case:
>> 
>> 1) create two separate playpen's of trunk.
>> 2) In one, unpack André's change that builds with Java 1.5
>> 3) ant build in that playpen
>> 4) and build in the other playpen
>> 
>> I observe that each time I switch playpen's, I am getting these
>> serialization errors, which implies to me the separate playpen's are
>> _sharing_ a cache.  That's clearly not right!
>> 
>> Questions:
>> 
>> a) Why would the caches be stored in the java tmpdir?
>> b) If we use the java tmpdir, don't we also have to make sure that we don't
>> collide with other apps (including our own) that might be using that dir?
>> 
>> On 2011-02-03, at 12:53, Henry Minsky wrote:
>> 
>>> Change hqm-20110203-9Nc by [email protected] on 2011-02-03 12:48:53
>> EST
>>>   in /Users/hqm/openlaszlo/trunk3
>>>   for http://svn.openlaszlo.org/openlaszlo/trunk
>>> 
>>> Summary: clear caches more thoroughly
>>> 
>>> New Features:
>>> 
>>> Bugs Fixed: LPP-9733
>>> 
>>> Technical Reviewer: ptw
>>> QA Reviewer: andre.bargull
>>> Doc Reviewer: (pending)
>>> 
>>> Documentation:
>>> 
>>> Release Notes:
>>> 
>>> Overview:
>>> 
>>> 
>>> Details:
>>> 
>>> some caches are stored in the java tmpdir, which we were not cleaning
>> with ant clean
>>> 
>>> 
>>> 
>>> Tests:
>>> 
>>> A little hard to reproduce, but when you upgrade to JAVACC 5.0, and do a
>> clean build, you should
>>> not get any cache errors
>>> 
>>> Files:
>>> M       build.xml
>>> 
>>> 
>>> Changeset:
>> http://svn.openlaszlo.org/openlaszlo/patches/hqm-20110203-9Nc.tar
>> 
>> 
> 
> 
> -- 
> Henry Minsky
> Software Architect
> [email protected]


Reply via email to